view Documentation/networking/ipvs-sysctl.txt @ 897:329ea0ccb344

balloon: try harder to balloon up under memory pressure.

Currently if the balloon driver is unable to increase the guest's
reservation it assumes the failure was due to reaching its full
allocation, gives up on the ballooning operation and records the limit
it reached as the "hard limit". The driver will not try again until
the target is set again (even to the same value).

However it is possible that ballooning has in fact failed due to
memory pressure in the host and therefore it is desirable to keep
attempting to reach the target in case memory becomes available. The
most likely scenario is that some guests are ballooning down while
others are ballooning up and therefore there is temporary memory
pressure while things stabilise. You would not expect a well behaved
toolstack to ask a domain to balloon to more than its allocation nor
would you expect it to deliberately over-commit memory by setting
balloon targets which exceed the total host memory.

This patch drops the concept of a hard limit and causes the balloon
driver to retry increasing the reservation on a timer in the same
manner as when decreasing the reservation.

Also if we partially succeed in increasing the reservation
(i.e. receive less pages than we asked for) then we may as well keep
those pages rather than returning them to Xen.

Signed-off-by: Ian Campbell <ian.campbell@citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Fri Jun 05 14:01:20 2009 +0100 (2009-06-05)
parents 831230e53067
line source
1 /proc/sys/net/ipv4/vs/* Variables:
3 am_droprate - INTEGER
4 default 10
6 It sets the always mode drop rate, which is used in the mode 3
7 of the drop_rate defense.
9 amemthresh - INTEGER
10 default 1024
12 It sets the available memory threshold (in pages), which is
13 used in the automatic modes of defense. When there is no
14 enough available memory, the respective strategy will be
15 enabled and the variable is automatically set to 2, otherwise
16 the strategy is disabled and the variable is set to 1.
18 cache_bypass - BOOLEAN
19 0 - disabled (default)
20 not 0 - enabled
22 If it is enabled, forward packets to the original destination
23 directly when no cache server is available and destination
24 address is not local (iph->daddr is RTN_UNICAST). It is mostly
25 used in transparent web cache cluster.
27 debug_level - INTEGER
28 0 - transmission error messages (default)
29 1 - non-fatal error messages
30 2 - configuration
31 3 - destination trash
32 4 - drop entry
33 5 - service lookup
34 6 - scheduling
35 7 - connection new/expire, lookup and synchronization
36 8 - state transition
37 9 - binding destination, template checks and applications
38 10 - IPVS packet transmission
39 11 - IPVS packet handling (ip_vs_in/ip_vs_out)
40 12 or more - packet traversal
42 Only available when IPVS is compiled with the CONFIG_IPVS_DEBUG
44 Higher debugging levels include the messages for lower debugging
45 levels, so setting debug level 2, includes level 0, 1 and 2
46 messages. Thus, logging becomes more and more verbose the higher
47 the level.
49 drop_entry - INTEGER
50 0 - disabled (default)
52 The drop_entry defense is to randomly drop entries in the
53 connection hash table, just in order to collect back some
54 memory for new connections. In the current code, the
55 drop_entry procedure can be activated every second, then it
56 randomly scans 1/32 of the whole and drops entries that are in
57 the SYN-RECV/SYNACK state, which should be effective against
58 syn-flooding attack.
60 The valid values of drop_entry are from 0 to 3, where 0 means
61 that this strategy is always disabled, 1 and 2 mean automatic
62 modes (when there is no enough available memory, the strategy
63 is enabled and the variable is automatically set to 2,
64 otherwise the strategy is disabled and the variable is set to
65 1), and 3 means that that the strategy is always enabled.
67 drop_packet - INTEGER
68 0 - disabled (default)
70 The drop_packet defense is designed to drop 1/rate packets
71 before forwarding them to real servers. If the rate is 1, then
72 drop all the incoming packets.
74 The value definition is the same as that of the drop_entry. In
75 the automatic mode, the rate is determined by the follow
76 formula: rate = amemthresh / (amemthresh - available_memory)
77 when available memory is less than the available memory
78 threshold. When the mode 3 is set, the always mode drop rate
79 is controlled by the /proc/sys/net/ipv4/vs/am_droprate.
81 expire_nodest_conn - BOOLEAN
82 0 - disabled (default)
83 not 0 - enabled
85 The default value is 0, the load balancer will silently drop
86 packets when its destination server is not available. It may
87 be useful, when user-space monitoring program deletes the
88 destination server (because of server overload or wrong
89 detection) and add back the server later, and the connections
90 to the server can continue.
92 If this feature is enabled, the load balancer will expire the
93 connection immediately when a packet arrives and its
94 destination server is not available, then the client program
95 will be notified that the connection is closed. This is
96 equivalent to the feature some people requires to flush
97 connections when its destination is not available.
99 expire_quiescent_template - BOOLEAN
100 0 - disabled (default)
101 not 0 - enabled
103 When set to a non-zero value, the load balancer will expire
104 persistent templates when the destination server is quiescent.
105 This may be useful, when a user makes a destination server
106 quiescent by setting its weight to 0 and it is desired that
107 subsequent otherwise persistent connections are sent to a
108 different destination server. By default new persistent
109 connections are allowed to quiescent destination servers.
111 If this feature is enabled, the load balancer will expire the
112 persistence template if it is to be used to schedule a new
113 connection and the destination server is quiescent.
115 nat_icmp_send - BOOLEAN
116 0 - disabled (default)
117 not 0 - enabled
119 It controls sending icmp error messages (ICMP_DEST_UNREACH)
120 for VS/NAT when the load balancer receives packets from real
121 servers but the connection entries don't exist.
123 secure_tcp - INTEGER
124 0 - disabled (default)
126 The secure_tcp defense is to use a more complicated state
127 transition table and some possible short timeouts of each
128 state. In the VS/NAT, it delays the entering the ESTABLISHED
129 until the real server starts to send data and ACK packet
130 (after 3-way handshake).
132 The value definition is the same as that of drop_entry or
133 drop_packet.
135 sync_threshold - INTEGER
136 default 3
138 It sets synchronization threshold, which is the minimum number
139 of incoming packets that a connection needs to receive before
140 the connection will be synchronized. A connection will be
141 synchronized, every time the number of its incoming packets
142 modulus 50 equals the threshold. The range of the threshold is
143 from 0 to 49.